Novel high resolution array signal processing algorithms

This paper considers the problem of source localization High resolution methods. These techniques are used in various domains of physics such as the underwater acoustics (sonar), and electromagnetic (radar), but also in other areas such as telecommunications, geophysics, seismic, biomedical, medical imaging, and radio astronomy, where it is often necessary to locate several sources (active or passive) at the same time, with very good resolution, in order to separate them even when they are very close. In the conventional high-resolution array processing, source localization is performed by means of the inversion of the spectral matrix between sensors, inversion can be difficult when the separation between the sources of interest and noisemakers is not clear. We propose original methods for locating sources which, by means of a suitable decomposition, prevent the inversion of the spectral matrix. These methods are on the triangular factorization of the spectral matrix product of two matrices (LU in one case and QR in the other). The essential property of these matrices U and R that the information on the energies are better organized than in the spectral matrix. Furthermore, this technique allows better dynamics to estimate the signal sources, that using the original spectral matrix. Furthermore, we propose two new estimators for source localization without knowing the number of sources a priori. The big advantage of these new estimators is to improve the localization in the presence of low signal to noise ratio (SNR). The performances of these estimators are comparable to that of the Multiple Signal Classification (MUSIC) algorithm which exhibit higher computational load and needs the knowledge of the number of sources. The new estimators decrease the computational complexity. Even if we use existing mathematical tools, this is the first time such an approach is proposed in array processing.